DRAFTVERSION DECEMBER 16, 2016 Preprint typeset using LATEX style AASTeX6 v. 1.0

ALMA OBSERVATIONS OF MASSIVE MOLECULAR GAS FILAMENTS ENCASING RADIO BUBBLES IN THE CLUSTER

H.R.RUSSELL1∗,M.MCDONALD2,B.R.MCNAMARA3,4,A.C.FABIAN1, P. E. J. NULSEN5,6,M.B.BAYLISS2,7,B.A. BENSON8,9,10,M.BRODWIN11,J.E.CARLSTROM10,9,A.C.EDGE12,J.HLAVACEK-LARRONDO13, D. P. MARRONE14,C.L. REICHARDT15 AND J.D.VIEIRA16 1 Institute of Astronomy, Madingley Road, Cambridge CB3 0HA 2 Kavli Institute for Astrophysics and Space Research, Massachusetts Institute of Technology, 77 Massachusetts Avenue, Cambridge, MA 02139, USA 3 Department of Physics and Astronomy, University of Waterloo, Waterloo, ON N2L 3G1, Canada 4 Perimeter Institute for Theoretical Physics, Waterloo, ON N2L 2Y5, Canada 5Harvard-Smithsonian Center for Astrophysics, 60 Garden Street, Cambridge, MA 02138, USA 6ICRAR, University of Western Australia, 35 Stirling Hwy, Crawley, WA 6009, Australia 7Department of Physics and Astronomy, Colby College, 5100 Mayflower Hill Dr, Waterville, ME 04901, USA 8Fermi National Accelerator Laboratory, Batavia, IL 60510-0500, USA 9Department of Astronomy and Astrophysics, University of Chicago, Chicago, IL 60637, USA 10Kavli Institute for Cosmological Physics, University of Chicago, Chicago, IL 60637, USA 11Department of Physics and Astronomy, University of Missouri, Kansas City, MO 64110, USA 12Department of Physics, Durham University, Durham DH1 3LE 13Département de Physique, Université de Montréal, Montréal, QC H3C 3J7, Canada 14Steward Observatory, University of Arizona, 933 North Cherry Avenue, Tucson, AZ 85721, USA 15School of Physics, University of Melbourne, Parkville VIC 3010, Australia 16Department of Astronomy and Department of Physics, University of Illinois, 1002 West Green St., Urbana, IL 61801, USA

ABSTRACT 10 We report new ALMA observations of the CO(3-2) line emission from the 2.1 ± 0.3 × 10 M molecular gas reservoir in the central of the Phoenix cluster. The cold molecular gas is fuelling a vigorous starburst at a −1 rate of 500−800M yr and powerful black hole activity in the form of both intense quasar radiation and radio jets. The radio jets have inflated huge bubbles filled with relativistic plasma into the hot, X-ray atmospheres surrounding the host galaxy. The ALMA observations show that extended filaments of molecular gas, each 10 − 20 kpc long with a mass of several billion solar masses, are located along the peripheries of the radio bubbles. The smooth velocity gradients and narrow line widths along each filament reveal massive, ordered molecular gas flows around each bubble, which are inconsistent with gravitational free-fall. The molecular clouds have been lifted directly by the radio bubbles, or formed via thermal instabilities induced in low entropy gas lifted in the updraft of the bubbles. These new data provide compelling evidence for close coupling between the radio bubbles and the cold gas, which is essential to explain the self-regulation of feedback. The very feedback mechanism that heats hot atmospheres and suppresses formation may also paradoxically stimulate production of the cold gas required to sustain feedback in massive . Keywords: cooling flows — galaxies:active — galaxies: clusters: Phoenix — radio lines: galaxies

1. INTRODUCTION galaxies and central cluster galaxies have also revealed huge arXiv:1611.00017v2 [astro-ph.GA] 15 Dec 2016 The energy output by active galactic nuclei (AGN) has cavities where the hot gas has been displaced by expanding long been recognized as sufficient to unbind the interstel- radio bubbles inflated by radio jets (McNamara et al. 2000; lar medium from even the most massive host galaxies (Silk Fabian et al. 2000, 2006). Known as AGN feedback, these & Rees 1998). Recent observations of ionized and molecu- energetic outbursts are therefore observed to couple effec- lar gas outflows driven by intense radiation or radio jet ac- tively to the cold and warm interstellar gas and the hot gas tivity from the central AGN show that this energy can be atmospheres surrounding massive galaxies. AGN feedback efficiently coupled to the surrounding interstellar gas (eg. is an essential mechanism in galaxy formation that powers Morganti et al. 2005; Nesvadba et al. 2006; Feruglio et al. gas outflows to truncate massive galaxy growth. This pro- 2010; Rupke & Veilleux 2011; Alatalo et al. 2011; Dasyra cess is thought to produce the observed evolution of galaxies & Combes 2011; Morganti et al. 2015). Chandra X-ray ob- from gas-rich, star forming systems to ‘red and dead’ ellipti- servations of the hot atmospheres surrounding giant elliptical cals and imprint the observed coevolution of massive galax- 2 ies and supermassive black holes (SMBHs; Magorrian et al. ture AGN activity. 1998; Croton et al. 2006; Bower et al. 2006). Here we present new ALMA observations of the CO(3- However, the details of how a SMBH can regulate the 2) emission from the molecular gas in the central galaxy growth of its host environment over many orders of magni- of the Phoenix cluster. Discovered with the South Pole tude in spatial scale are still poorly understood. In the most Telescope, the Phoenix cluster (SPT-CLJ2344-4243), at red- massive galaxies at the centres of cool core galaxy clusters, shift z = 0.596, is the most luminous X-ray cluster known the radiative cooling time of the hot gas atmosphere can fall (Williamson et al. 2011; McDonald et al. 2012), and the −1 below a Gyr and heat input from the AGN must be distributed 500 − 800M yr starburst hosted by its central galaxy is throughout the central 100 kpc to prevent the formation of a amongst the largest found in any galaxy below 1. cooling flow (eg. Edge et al. 1992; Peres et al. 1998; Voigt The star formation is observed in bright filaments stretch- & Fabian 2004). Without this energy input, gas would cool ing beyond 100 kpc, sustained by a 20 billion solar mass unimpeded from the cluster atmosphere and produce at least reservoir of molecular gas (McDonald et al. 2013a, 2014). 12 an order of magnitude more molecular gas and star formation The stellar mass of the massive central galaxy is 3×10 M than is observed in central cluster galaxies (Johnstone et al. (McDonald et al. 2012, 2013a) and it hosts an unusual cen- 1987; Edge 2001; Salomé & Combes 2003). Radio jets pow- tral that is powering both intense ered by the central AGN inflate buoyant radio bubbles and radiation and relativistic jets. Observations show these to be drive shocks and sound waves into the distinct modes of AGN feedback. The black hole may be to produce distributed heating throughout the cluster core (for in the process of transitioning from a radiatively powerful reviews see McNamara & Nulsen 2007, 2012; Fabian 2012). quasar to a radio galaxy (eg. Churazov et al. 2005; Russell X-ray studies of large samples of galaxy groups and clusters et al. 2013; Hlavacek-Larrondo et al. 2013) whose mechani- show that this energy input is sufficient to replace the major- cal power output of ∼ 1046 erg s−1 is among the largest mea- ity of the radiative losses from the cluster gas on large scales sured (eg. Hlavacek-Larrondo et al. 2015; McDonald et al. (Bîrzan et al. 2004; Dunn & Fabian 2006; Rafferty et al. 2015). The Phoenix cluster therefore hosts an extreme exam- 2006). The heating rate supplied by the AGN is also observed ple of this common mechanism in galaxy evolution. Both the to be closely correlated with the cooling rate of the cluster powerful black hole activity and the vigorous starburst are atmosphere, which implies a highly effective feedback loop fuelled by the massive cold molecular gas reservoir, whose operating over this huge range of spatial scales. A few per structure can now be resolved with ALMA to understand how cent of the most rapidly cooling cluster gas does cool to low these processes are regulated. temperatures and likely feeds the observed cold molecular We assume a standard ΛCDM cosmology with H0 = −1 −1 gas reservoirs and star formation in the central galaxy. Al- 70 km s Mpc , ΩM = 0.27 and ΩΛ = 0.73. At the redshift though the level of gas cooling falls far below the predictions of the Phoenix cluster z = 0.596 (Ruel et al. 2014; McDonald of cooling flows, prompt accretion of this residual compo- et al. 2015), 1 arcsec corresponds to 6.75 kpc. nent is likely required to link the large scale cooling rate to the energy output of the AGN in an efficient feedback loop. 2. DATA REDUCTION Observations of ionized and molecular gas at the centres of The brightest cluster galaxy (BCG) in the Phoenix cluster clusters have revealed cool gas filaments extending radially was observed by ALMA on 15 and 16 June 2014 (Cycle 2, from the galaxy centre towards radio bubbles inflated by the ID 2013.1.01302.S; PI McDonald) simultaneously covering jet (Fabian et al. 2003; Salomé et al. 2006, 2008; Hatch et al. the CO(3-2) line at 216.66 GHz and the sub-mm continuum 2006; Lim et al. 2008). In the , the velocity emission in three additional basebands at 219.5, 230.5 and structure of the Hα-emitting filaments, which are coincident 232.5 GHz. The single pointing observations were centred with detections of CO emission from the IRAM 30 m tele- on the nucleus with a field of view of 28.5 arcsec. The total scope, traces streamlines underneath a buoyantly rising radio time on source was 58.5min split into nine ∼ 6min obser- bubble (Salomé et al. 2006, 2011). ALMA observations of vations and interspersed with observations of the phase cali- molecular gas at the centres of clusters (Russell et al. 2014, brator J2357-5311. This bright quasar was also observed for 2016; McNamara et al. 2014; David et al. 2014; Tremblay bandpass and flux calibration. The observations utilised 35 et al. 2016; Vantyghem et al. 2016) have shown cold gas fila- antennas with baselines of 20 − 650 m. The frequency divi- ments extending along the trajectories of radio bubbles. The sion correlator mode was used for the spectral line observa- molecular clouds have either been lifted directly by the bub- tion with a 1.875 GHz bandwidth and frequency resolution bles or cooled in situ from warmer, thermally unstable gas of 7812 kHz. The velocity channels were binned to a res- lifted in their wakes. The velocities of the molecular clouds olution of 12 km s−1 for the subsequent analysis. Based on are remarkably slow compared to the stellar velocity disper- optical spectroscopy of the central galaxy, we use a velocity sion in these massive galaxies and lie well below the galaxy’s center at a redshift z = 0.596, which also corresponds to the escape velocity. The molecular gas will likely fall back to- velocity center of the molecular emission peak. We note that wards the galaxy centre and fuel both star formation and fu- optical IFU observations have revealed a very dynamic envi- 3

Figure 1. Left: Phoenix CO(3-2) integrated intensity map for velocities −430 to +600 km s−1 covering both the central gas peak and the extended filaments. Contour levels are 2σ, 4σ, 6σ, 8σ, 10σ, 15σ etc, where σ = 0.067 Jy/beam.km s−1 . The ALMA beam is shown lower left and the continuum point source location is shown by the black cross. Right: Phoenix CO(3-2) spectrum for a 600 × 600 region centred on the nucleus. The best fit model is shown by the solid black line and individual Gaussian components are shown by the dashed lines (see Table1).

Figure 2. Left: CO(3-2) integrated intensity map for velocities 0 to +480 km s−1 covering the extended filaments. Contour levels are at 2σ, 4σ, 6σ, 8σ, 10σ, 15σ, 20σ etc., where σ = 0.042 Jy/beam.km s−1 . The ALMA beam is shown in the lower left corner (0.60 arcsec × 0.56 arcsec). The X-ray cavities are shown by the dashed white contours, which correspond to the negative residuals after a smooth model was subtracted from the X-ray surface brightness (McDonald et al. 2015). Right: HST image combining the F475W (blue), F625W (green) and F814W (red) filters (McDonald et al. 2013a). Both images cover the same field of view. ronment in the ionized gas and bulk motions could produce image quality. Different weightings were explored to de- a systematic offset in the gas velocities with respect to the termine the optimum for imaging. Natural weighting de- gravitational potential of the BCG (McDonald et al. 2014). tected the extended filaments at the highest signal-to-noise The observations were calibrated in CASA version 4.3.1 but no major differences could be discerned between the var- (McMullin et al. 2007) using the ALMA pipeline reduc- ious weightings due to the good uv coverage. The final data tion scripts. Continuum-subtracted data cubes were gen- cube used natural weighting and had a synthesized beam of erated using CLEAN and UVCONTSUB. Additional self- 0.60 arcsec × 0.56 arcsec with P.A. = −37.9deg. The rms calibration did not produce significant improvement in the noise in the line-free channels was 0.3 mJy/beam at CO(3- 4

2) for 12 km s−1 channels. An image of the continuum emis- largest projected distance around the bubble, which could ex- sion with an rms noise of 0.019 mJy/beam was generated plain the remarkable coincidence between the filaments and by combining spectral channels from all four basebands that the bubble edges and the limited amount of molecular gas were free of line emission. The continuum image was pro- projected across either bubble. duced using natural weighting and the synthesized beam was The total CO(3-2) intensity of 8.9±1.5 Jy km s−1 is almost 0.59 arcsec × 0.53 arcsec with P.A. = −48.7deg. a factor of 2 greater than the integrated intensity measured by the SMA of 5.3 ± 1.4 Jy km s−1 (McDonald et al. 2014). The SMA observation was taken at very low elevation, is 3. RESULTS only modestly significant and likely affected by a substan- 3.1. Molecular gas morphology tial continuum subtraction uncertainty, making it particularly The CO(3-2) molecular line emission peaks at the galaxy difficult to calibrate. The measured FWHM of ∼ 400 km s−1 center, offset by ∼ 0.3 arcsec to the W of the radio nucleus from the SMA observation is also significantly less than de- (Fig.1 left). The central molecular gas peak extends along termined the FWHM of ∼ 670±20 km s−1 for a single com- a NE to SW axis across the nucleus. Two filaments extend ponent spectral fit to the ALMA total spectrum. This dis- 3 − 4 arcsec (20 − 27 kpc) to the NW and SE of the central crepancy in the total CO(3-2) flux could therefore be due to emission peak. The emission also extends for several arcsec uncertainty in the SMA continuum subtraction. as a broader structure to the S of the nucleus. Fig.1 (right) shows the continuum-subtracted total CO(3-2) spectral line 3.2. Velocity structure profile extracted from a 600 × 600 region centred on the nu- The velocity structure of the molecular gas was mapped cleus and covering all extended emission. The line profile is by extracting spectra in synthesized beam-sized regions each very broad, covering ∼ 1000 km s−1 , and consists of multi- centred on each spatial pixel in the ALMA cube. The ex- ple velocity components. The total spectrum was fitted with tracted spectra were fitted with MPFIT using one, two or three three Gaussian components using the package MPFIT (Mark- Gaussian components. We required at least 3σ significance wardt 2009). The brightest velocity component is centred for the detection of an emission line in each region based on the galaxy’s systemic velocity and has the largest FWHM on 1000 Monte Carlo simulations of the spectrum. Figs.3 of 450 ± 80 km s−1 . At least two additional velocity compo- and4 show the maps of the line centre and FWHM that were nents are redshifted to 310 ± 20 km s−1 and 620 ± 30 km s−1 generated for each best-fit velocity component. These maps and have significantly lower FWHMs of ∼ 250 km s−1 . The show two distinct components in the molecular gas: the ex- best fit results, corrected for primary beam response and in- tended filaments characterized by smooth velocity gradients strumental broadening, are shown in Table1. and narrow FWHM= 100 − 200 km s−1 and a compact cen- The most luminous redshifted component covering the ve- tral gas peak with much broader FWHM= 300−550 km s−1 . locities from ∼ 0 to ∼ 480 km s−1 corresponds to the most All three extended filaments have ordered velocity gradi- extended emission (Fig.2). The remaining molecular gas ents along their lengths and low FWHM< 250 km s−1 de- at > 500 km/s and < 0 km s−1 lies within 1.5 arcsec of the creasing to < 150 km s−1 at the largest radii in the NW and S nucleus. Fig.2 clearly shows that this velocity component filaments. The smooth velocity gradients and narrow FWHM traces the most extended emission from the NW and SE fil- over the length of each extended structure reveal a steady, or- aments and the third filament to the S. The filament widths dered flow of molecular gas around and beneath the radio are unresolved and may consist of many individual strands bubbles. The velocities at the furthest extent of each filament (Fabian et al. 2008). Each filament coincides with regions of are similar at ∼ 250 km s−1 and increase towards the galaxy bright ionized gas emission, dust and filamentary star forma- nucleus with the highest velocities at the smallest radii. In tion previously detected in optical and UV observations (Fig. the SE filament, the velocity increases with decreasing radius 2 right; McDonald et al. 2013a, 2014). to ∼ 600 km s−1 to the E of the nucleus. The steady veloc- The filaments are draped around the rims of two large X- ity gradient suggests that the SE filament forms a continuous ray cavities (Fig.2, left) detected in deep Chandra X-ray ob- structure to the galaxy centre. The velocity gradient along the servations (McDonald et al. 2015). These cavities are each NW filament, from +280 km s−1 at 2.3 arcsec radius (16 kpc) 9 − 15 kpc across and centered at a radius of ∼ 17 kpc and to 0 km s−1 at the nucleus, also indicates a separate, contin- were created as radio bubbles inflated by the AGN displaced uous structure. The S filament is fainter, shorter and wider the cluster gas. The SE and S filaments encase the inner and has a shallower velocity gradient. The FWHM is below half of the southern X-ray cavity, which is larger and has the 100 km s−1 at 2.5 arcsec radius (17 kpc) but quickly broad- greater cavity power of the two. The NW filament lies along ens to > 200 km s−1 with decreasing radius indicating that the W edge of the northern X-ray cavity but no significant the S filament is disrupted towards the galaxy centre. The SE counterpart is detected along the E edge. The filaments may and NW filaments meet in projection at the nucleus. The ve- form part of a patchy thin shell surrounding the inner half of locity differential is large across the nucleus at +600 km s−1 each bubble. This molecular shell would be brightest at the to 0 km s−1 with no evidence for disruption at this resolution 5

Table 1. Fit parameters for the total CO(3-2) spectrum shown in Fig.1 corrected for primary beam response.

Region χ2/dof Component Integrated intensity Peak FWHM Velocity shift (Jy km/s) (mJy) (km/s) (km/s) Total 198/141 1 5.3 ± 1.0 11.0 ± 1.0 450 ± 80 20 ± 50 2 2.3 ± 1.0 8.5 ± 3.0 260 ± 70 310 ± 20 3 1.3 ± 0.4 4.9 ± 0.6 250 ± 60 630 ± 30

Figure 3. Velocity line centre for each component. Spectral fitting reveals that the molecular gas structure has two distinct velocity components. The first component (left) traces the filaments and has a narrow FWHM∼ 100 − 200 km s−1 and smooth velocity gradients along their lengths. The second component (right) corresponds to the central gas peak and has a much higher FWHM∼ 300 − 550 km s−1 , lower velocities and a gradient E-W across the nucleus. The contours correspond to Fig.1 (left)

Figure 4. Same as Fig.3 but showing the FWHM of each velocity component. and no significant increase in FWHM of the velocity compo- ent from +120 km s−1 to −80 km s−1 across the nucleus (Fig. nents for each filament. Collisions between the gas clouds in 3). The velocity gradient across the nucleus appears to lie these two ordered filaments may produce a sharp transition perpendicular to the velocity gradients along the NW and SE in velocity to the second observed velocity component. filaments. The FWHM peaks along the projected intersection The central compact molecular gas peak forms a separate of the two filaments. The increase in FWHM could indicate velocity structure from the extended filaments with a much variation in the orientation of the filaments at the galaxy cen- higher FWHM= 300−550 km s−1 and a E-W velocity gradi- tre but, although it could be a contributing effect, this would 6 require a reversal of the velocity gradient in a similar length spatial resolution observations would be required to deter- of filament oriented along the line of sight at the galaxy cen- mine if this is a disk. tre. It is more likely that the FWHM is intrinsically higher in 3.3. Molecular gas mass the central molecular gas peak. The E-W velocity gradient centred on the nucleus is consistent with ordered motion or By assuming a CO-to-H2 conversion factor XCO and a rotation about the AGN within a radius of ∼ 1 arcsec (7 kpc) line ratio CO(3-2)/CO(1-0) ∼ 0.8 (Edge 2001; Russell et al. but higher spatial resolution observations are required to de- 2016), the total molecular gas mass can be inferred from the termine if this corresponds to a disk. integrated CO intensity: The strong velocity gradients in the molecular gas are    2 comparable to those observed for the warm ionized gas 4 1 SCO∆ν DL Mmol = 1.05×10 XCO M , (McDonald et al. 2014). The ionized gas, traced by the 1 + z Jy km s−1 Mpc [OIII]λ4959, 5007 doublet, shows a relatively smooth gra- (1) −1 dient from +700 km s to the SE of the nucleus decreasing where DL is the luminosity distance, z is the redshift of −1 −1 to −400 km s to the NW. The velocity is ∼ 0 km s at the the BCG and SCO∆ν is the integrated CO(1-0) intensity. peak of the [OII] emission around the nucleus. The low ve- However, the molecular gas mass is particularly sensitive −1 −1 locities of +200 km s to −200 km s at the centre corre- to the CO-to-H2 conversion factor, which is quite uncer- spond to the bright central peak and are consistent with the tain and not expected or observed to be universal (see Bo- gas velocities around the nucleus in the molecular gas. The latto et al. 2013 for a review). Previous studies of BCGs high velocities in the ionized gas to the SE and NW are com- in cool core clusters (Edge 2001; Salomé & Combes 2003; parable to the bright, innermost regions of the SE and NW Russell et al. 2014; McNamara et al. 2014) have used the 20 −2 −1 −1 filaments. Although the ionized gas velocities appear to de- Galactic value XCO = 2×10 cm ( K km s ) (Solomon crease at larger radii, which corresponds to the outer regions et al. 1987; Solomon & Vanden Bout 2005). However in of the filaments, the emission is faint and extends beyond the intense starburst galaxies and ULIRGs such as the BCG in field of view (McDonald et al. 2014). the Phoenix cluster, the molecular gas exists at higher densi- The fraction of the total CO(3-2) flux in each filament was ties and temperatures producing an extended warm gas phase estimated by using the velocity structure to separate the fila- with a much higher column density than a quiescent sys- ments from the compact central emission peak. Using a spec- tem. Under these conditions, the CO emission is more lu- trum extracted from a region covering the central peak, we minous and XCO should be lowered (eg. Downes et al. 1993; determined the integrated flux in each filament based on the Iono et al. 2007; Aravena et al. 2016). The high star for- −1 −2 best-fit model for their distinct velocity structures. This was mation density of 5M yr kpc within the central 10 kpc added to the flux determined from conservative regions cov- (McDonald et al. 2014), warm dust temperature of 87 K ering only the extended structure of each filament. The low and large FWHM in the galaxy center suggest that a lower velocity structure at large radius in each filament could not XCO is appropriate for the central molecular gas structure be easily spectrally separated from the low velocities of the and potentially also for the filaments. We therefore assume 20 −2 −1 −1 gas across the nucleus. Therefore it was not possible to use XCO = 0.4 × 10 cm ( K km s ) (Downes & Solomon a purely spectral or purely spatial separation of the extended 1998) but note that Chandra observations measure subso- and compact structures. From this hybrid technique, we es- lar metallicity of 0.5Z in the surrounding ICM. Low metal timate that the SE filament contains ∼ 25% of the total flux abundance likely will boost XCO over our assumed value, un- and the NW filament contains ∼ 15%. The extent of the S fil- less the cool gas in the filaments has an increased metallicity ament is particularly uncertain. Based on the assumption that over the ambient medium (eg. Panagoulia et al. 2013). all the emission to the S of the nucleus with a FWHM below For the integrated CO(3-2) intensity of 8.9±1.5 Jy km s−1 , −1 10 250 km s is associated with the S filament then it contains the total molecular gas mass is 2.1 ± 0.3 × 10 M . As dis- ∼ 10 − 15% of the total flux. cussed in section 3.1, the integrated intensity is almost a fac- In summary, roughly half of the total flux lies in three tor of two higher than that found by the earlier SMA obser- extended filaments which have ordered velocity gradients vation (McDonald et al. 2014), which is likely due to un- along their lengths and low FWHM< 250 km s−1 . The ve- certainty in the continuum subtraction for the SMA result. locities at the furthest extent of each filament are similar at Note also that McDonald et al.(2014) assume a line ratio ∼ 250 km s−1 and increase towards the galaxy nucleus with CO(3-2)/CO(1-0) ∼ 0.5, which produces a similar molecular the highest velocities at the smallest radii. The central com- gas mass despite the difference in integrated intensity. The pact emission peak forms a separate velocity structure with a central molecular gas peak in the Phoenix cluster accounts much higher FWHM= 300 − 550 km s−1 and a velocity gra- for ∼ 50% of the total CO flux and therefore has a molec- −1 10 dient lying perpendicular to the filaments from +120 km s ular gas mass of 1.0 ± 0.2 × 10 M . The SE, NW and −1 10 10 to −80 km s across the nucleus. The velocity structure is S filaments contain ∼ 0.5 × 10 M , ∼ 0.3 × 10 M and 10 consistent with ordered motion around the nucleus but higher ∼ 0.3 × 10 M , respectively. The uncertainty on the XCO 7 factor increases the uncertainty on the molecular gas masses preferentially host cold molecular gas in excess of several 9 to roughly a factor of a few. This estimate of the molecular 10 M (Edge 2001; Salomé & Combes 2003). The molec- gas mass also appears low when compared with correlations ular gas structures are observed to be coincident with bright with the Hα luminosity (Salomé & Combes 2003) and the optical emission line nebulae and the most rapidly cooling dust-to-gas ratio of ∼ 20 (McDonald et al. 2012). However, X-ray gas (Fabian et al. 2003; Salomé et al. 2006). In the our conclusions are not qualitatively altered by the estimated Phoenix cluster, the molecular filaments are similarly coin- uncertainty. cident with the brightest, soft X-ray emission, ionized gas plumes and dust regions. The X-ray gas cooling rate mea- 3.4. AGN continuum +340 −1 sured with XMM-Newton RGS of 120−120 M yr (Tozzi An unresolved central continuum source was detected at et al. 2015) is consistent with the observed mass of molec- RA 23:44:43.902, Dec -42:43:12.53 with a flux of 2.5 ± ular gas originating in cooling of the hot atmosphere over 0.1 mJy at 225.09 GHz. The position and flux are consis- roughly the buoyant rise time of the inner and outer bubbles tent with the SMA continuum detection at 3 mJy (McDonald in the Phoenix cluster (50 − 120 Myr, see also Russell et al. et al. 20141). The ALMA continuum image does not reveal 2014; McNamara et al. 2014). An alternative merger origin any extended emission due to star formation. The contin- for such a substantial mass of molecular gas would require uum source is coincident with radio and hard X-ray point multiple gas-rich donor galaxies, which are rare at the cen- source emission. Low frequency radio observations from the tres of rich clusters (eg. Young et al. 2011). It is more likely SUMSS and ATCA archives suggest a synchrotron contin- that the molecular gas originated in gas cooling from the sur- uum slope of S ∝ ν−1.35 and therefore we expect synchrotron rounding hot atmosphere. emission from the AGN of ∼ 0.04 mJy at 220 GHz (McDon- The molecular gas clouds have either been subsequently ald et al. 2014). The observed point source flux is therefore lifted into extended filaments by the expanding radio bubbles consistent with a combination of synchrotron emission and or formed in filaments via thermal instabilities induced in up- dust emission from the SMBH’s immediate environment. lifted low entropy X-ray gas. Radio jets have been observed to drive significant outflows of molecular gas from galax- 4. DISCUSSION ies (Morganti et al. 2005; Dasyra & Combes 2012; Morganti In the central galaxy of the Phoenix cluster, massive et al. 2015). For the Phoenix cluster, McDonald et al.(2015) molecular gas filaments form dense, cold rims around both estimated the mechanical jet energy from the work done dis- of the inner X-ray cavities, where hot gas has been displaced placing the hot gas against the surrounding pressure. By mea- by radio jet activity. These observations now clearly demon- suring the size of the inner cavities and the local gas pressure, 59 strate that the structure of the largest molecular gas reservoirs the cavity enthalpy was estimated as 4.4 − 6.7 × 10 erg. located in the most massive galaxies is shaped by the expan- Only a few per cent of this mechanical energy could sup- sion and trajectory of the radio bubbles. Previous sub-mm ply the observed kinetic energy of the molecular filaments. observations of brightest cluster galaxies have indicated ten- However, the coupling mechanism between dense, molecu- tative correlations between X-ray cavity axes and the orien- lar clouds and radio bubbles is unclear and this mechanism tations of molecular gas filaments, including ALMA obser- would have to be extremely efficient to lift 50% of the cold vations of Abell 1835 and PKS 0745-191 (McNamara et al. gas into extended filaments. Volume-filling X-ray gas would 2014; Russell et al. 2016). IRAM observations of the nearby be much easier to lift. Perseus cluster detected molecular gas coincident with re- Outflows of hot X-ray gas, enriched with metals by stellar gions of the complex optical emission line nebula, includ- activity in the central galaxy, are observed in galaxy clus- ing several filaments of ionized gas that extend toward radio ters as plumes of high metallicity gas lifted along the jet axis bubbles (Salomé et al. 2006; Lim et al. 2008; Salomé et al. for tens to hundreds of kpc (Simionescu et al. 2008; Kirk- 2011). These observed direct interactions between the cold patrick et al. 2009). Low entropy X-ray gas should become gas, which fuels the starburst and black hole activity, and the thermally unstable when lifted to a radius where its cooling jet-blown bubbles are essential to explain the observed close time approaches the infall time (Nulsen 1986; Pizzolato & regulation of AGN feedback. Soker 2005; McNamara et al. 2014, 2016). Theoretical mod- 10 els further indicate that lifting low entropy gas in the updraft The total molecular gas mass of 2.1 ± 0.3 × 10 M (sec- tion 3.3) in the central galaxy of the Phoenix cluster is sub- of rising radio bubbles stimulates condensation of molecular stantially higher than that typically found in early-type galax- clouds (Li & Bryan 2014; Brighenti et al. 2015; Voit et al. ies (Young et al. 2011). BCGs situated in dense cluster at- 2016). Therefore, an infall time that is significantly longer mospheres with short radiative cooling times are known to than the free-fall time, will enhance thermal instabilities and promote the formation of molecular gas clouds in the bub- bles’ wakes (McNamara et al. 2016). 1 Note that the lower SMA continuum flux given in McDonald et al. 2014 The inner radio bubbles in the Phoenix cluster displace 10 is a typographical error. ∼ 3 − 5 × 10 M and therefore, by Archimedes’ principle, 8

Figure 5. Left: Integrated intensity map for velocities −430 to +600 km s−1 covering both the central gas peak and the extended filaments. Contours are 2σ, 4σ, 6σ, 8σ, 10σ, 15σ, 20σ, 25σ and 30σ, where σ = 0.065 Jy/beam.km s−1 . The white box shows the extraction region for the position-velocity diagram. Right: Position-velocity diagram for the SE to NW axis along the two brightest filaments summed over roughly the width of the synthesized beam. Model velocity profiles are shown by the dashed lines for gravitational free-fall with inclinations 8, 16 and 45◦. For the SE filament, initial radii are 15, 15 and 21 kpc, respectively. For the NW filament, initial radii are 11, 11 and 15 kpc, respectively. Inclinations < 15◦ are required to match the velocity gradient of both filaments.

10 could lift the 1.0 × 10 M of gas required to supply the this gravitational potential is given by molecular gas in the filaments. The similarity in the molec- ular gas velocity at large radius, and in the velocity gradi- s   2 1 1 ents beneath bubbles with apparently different dimensions, v(r) = v(r0) + 2GM − , (2) r + a r0 + a supports this scenario where the molecular gas cools and de- couples from the hot atmosphere and falls toward the galaxy where a is the scale radius (Re ∼ 1.8153a), v(r0) is the initial −1 center. Whilst the velocity range covers ∼ 1000 km s at velocity of the gas blob and r0 is its initial radius. McDon- the galaxy center, the molecular gas velocities at the outer tip ald et al.(2012) (see also McDonald et al. 2013a) determine of each filament, separated by ∼ 30 kpc, are consistent with an effective radius of ∼ 17 kpc from HST imaging in five ∼ 250 km s−1 . Such similar velocities suggest that this re- photometric bands. The scale radius is therefore estimated at mote molecular gas could be coupled to the hot atmosphere 9.4 kpc. (Hitomi Collaboration et al. 2016), which is moving relative The total galaxy mass was estimated from Chandra obser- to the BCG. Bulk motion of the cluster gas could also explain vations assuming hydrostatic equilibrium for the hot X-ray the bubble asymmetry. The smoothly increasing gas veloc- atmosphere in the gravitational potential. We used spectra ities with decreasing radius along the NW and SE filaments extracted in concentric annuli (McDonald et al. 2015) and indicate massive gas flows underneath the radio bubbles. Al- the NFWMASS model (Nulsen et al. 2010) in XSPEC (ver- though the velocity structure cannot cleanly distinguish be- sion 12.9.0; Arnaud 1996) to determine the best-fit NFW pro- tween inflow and outflow, the remarkable similarity in the file parameters assuming a spherical, hydrostatic atmosphere. molecular gas velocities at large radii suggests the molecu- +40 The best-fit scale radius rs = 200−30 kpc and the normaliza- 6 −3 lar gas could be decoupling from the hot atmosphere and the tion constant ρ0 = 7 ± 2 × 10 M kpc . We therefore es- 13 smoothly increasing velocities toward the galaxy centre sug- timate a galaxy total mass of ∼ 2 × 10 M within a radius gest subsequent infall. of ∼ 50 kpc. For the best-fit NFW profile, the cluster mass 14 The smooth velocity gradients along the SE and NW fil- within r500 ∼ 1.3 Mpc is M500 = 7 ± 4 × 10 M . This is aments are shown clearly in Fig.5. Following Lim et al. consistent with the mass determined from scaling relations (2008), we assume a Hernquist model for the gravitational with YSZ (Williamson et al. 2011) and with YX (McDonald potential of the central galaxy (Hernquist 1990) constrained et al. 2015). by parameters for the total galaxy mass M and effective ra- The remaining free parameters for the Hernquist model are dius Re. The velocity acquired by a gas blob that free-falls in the initial radius of the gas blob, the inclination of the trajec- tory to the line of sight and the initial velocity, where the clouds are coupled to the hot atmosphere. As discussed in 9 section 3.2, the outermost velocities in the filaments are sim- radio-jet activity. The molecular gas may have been directly ilar and therefore all models used 250 km s−1 . The initial lifted in the bubble wakes or formed in situ at large radius radius was selected to match each model to the outermost from uplifted low entropy X-ray gas that became thermally region of the appropriate filament. Fig.5 shows the PV dia- unstable. The gas velocities appear too low for the bulk of gram for the NW to SE axis along the NW and SE filaments the cold gas to escape the galaxy and the gas will eventu- with free-fall models for several inclination angles. The ob- ally fall back toward the galaxy center to feed the central gas served shallow velocity gradients can only be matched by the peak. The observed close coupling between the radio bubbles lowest inclination angles θ < 15◦, for which both filaments and the cold gas is essential to explain the self-regulation of are oriented close to the plane of the sky. Beyond the initial feedback and understand the stability of this mechanism in acceleration, the model velocity gradient does not depend on clusters over at least half the age of the universe (McDonald the initial radius and velocity. Although the inclination and et al. 2013b; Ma et al. 2013; Hlavacek-Larrondo et al. 2015). the total mass are degenerate, the total mass would have to be overestimated by at least a factor of a few to allow inclina- ACKNOWLEDGEMENTS tion angles of > 20◦. This would also likely require the total HRR and ACF acknowledge support from ERC Ad- 12 stellar mass of 3 × 10 M (McDonald et al. 2012, 2013a) vanced Grant Feedback 340442. MM acknowledges support to have been significantly overestimated. We therefore sug- by NASA through contracts HST-GO-13456 (Hubble) and gest that such stringent requirements for the orientations of GO4-15122A (Chandra). BRM acknowledges support from all three filaments, and similar results from ALMA obser- the Natural Sciences and Engineering Council of Canada vations of PKS 0745-191 and Abell 1835 (McNamara et al. and the Canadian Space Agency Space Science Enhancement 2014; Russell et al. 2016), demonstrate that the gas veloc- Program. PEJN acknowledges support from NASA contract ities are more likely intrinsically low. Rather than require NAS8-03060. BB is supported by the Fermi Research Al- that the observed filaments are all aligned in the plane of the liance, LLC under Contract No. De-AC02-07CH11359 with sky, we suggest that the filament velocities are inconsistent the United States Department of Energy. ACE acknowl- with gravitational free-fall. In addition to the effect of in- edges support from STFC grant ST/L00075X/1. JHL ac- clination, the infalling gas blobs are slowed, potentially by knowledges support from the Natural Sciences and Engi- magnetic tension (Fabian et al. 2008; Russell et al. 2016) or neering Council of Canada, the Canada Research Chairs pro- by cloud-cloud collisions within the central few kpc (Pizzo- gram and the Fonds de recherche Nature et technologies. CR lato & Soker 2005; Gaspari et al. 2015). acknowledges support from the Australian Research Coun- cil’s Discovery Projects funding scheme (DP150103208). We thank the reviewer for constructive comments and HRR 5. CONCLUSIONS thanks Adrian Vantyghem for helpful discussions. This paper 10 Half of the 2.1 ± 0.3 × 10 M molecular gas reservoir makes use of the following ALMA data: ADS/JAO.ALMA at the center of the Phoenix cluster lies in extended fila- 2013.1.01302.S. ALMA is a partnership of ESO (represent- ments draped around expanding radio bubbles inflated by rel- ing its member states), NSF (USA) and NINS (Japan), to- ativistic jets and powered by the SMBH. The filaments have gether with NRC (Canada), NSC and ASIAA (Taiwan), and smooth velocity gradients along their lengths and narrow line KASI (Republic of Korea), in cooperation with the Repub- widths consistent with massive, ordered gas flows around the lic of Chile. The Joint ALMA Observatory is operated by radio bubbles. Although the velocity structure alone does not ESO, AUI/NRAO and NAOJ. The scientific results reported allow us to distinguish cleanly between inflow or outflow, the in this article are based on data obtained from the Chandra massive molecular gas flow is clearly shaped by the recent Data Archive.

REFERENCES

Alatalo, K., Blitz, L., Young, L. M., et al. 2011, ApJ, 735, 88 —. 2012, A&A, 541, L7 Aravena, M., Spilker, J. S., Bethermin, M., et al. 2016, MNRAS, 457, 4406 David, L. P., Lim, J., Forman, W., et al. 2014, ApJ, 792, 94 Arnaud, K. A. 1996, in Astronomical Society of the Pacific Conference Downes, D., & Solomon, P. M. 1998, ApJ, 507, 615 Series, Vol. 101, Astronomical Data Analysis Software and Systems V, Downes, D., Solomon, P. M., & Radford, S. J. E. 1993, ApJL, 414, L13 ed. G. H. Jacoby & J. Barnes, 17–+ Dunn, R. J. H., & Fabian, A. C. 2006, MNRAS, 373, 959 Bîrzan, L., Rafferty, D. A., McNamara, B. R., Wise, M. W., & Nulsen, Edge, A. C. 2001, MNRAS, 328, 762 P. E. J. 2004, ApJ, 607, 800 Edge, A. C., Stewart, G. C., & Fabian, A. C. 1992, MNRAS, 258, 177 Bolatto, A. D., Wolfire, M., & Leroy, A. K. 2013, ARA&A, 51, 207 Fabian, A. C. 2012, ARA&A, 50, 455 Bower, R. G., Benson, A. J., Malbon, R., et al. 2006, MNRAS, 370, 645 Fabian, A. C., Johnstone, R. M., Sanders, J. S., et al. 2008, Nature, 454, 968 Brighenti, F., Mathews, W. G., & Temi, P. 2015, ApJ, 802, 118 Fabian, A. C., Sanders, J. S., Crawford, C. S., et al. 2003, MNRAS, 344, Churazov, E., Sazonov, S., Sunyaev, R., et al. 2005, MNRAS, 363, L91 L48 Croton, D. J., Springel, V., White, S. D. M., et al. 2006, MNRAS, 365, 11 Fabian, A. C., Sanders, J. S., Taylor, G. B., et al. 2006, MNRAS, 366, 417 Dasyra, K. M., & Combes, F. 2011, A&A, 533, L10 Fabian, A. C., Sanders, J. S., Ettori, S., et al. 2000, MNRAS, 318, L65 10

Feruglio, C., Maiolino, R., Piconcelli, E., et al. 2010, A&A, 518, L155 Morganti, R., Oosterloo, T., Oonk, J. B. R., Frieswijk, W., & Tadhunter, C. Gaspari, M., Brighenti, F., & Temi, P. 2015, A&A, 579, A62 2015, A&A, 580, A1 Hatch, N. A., Crawford, C. S., Johnstone, R. M., & Fabian, A. C. 2006, Morganti, R., Tadhunter, C. N., & Oosterloo, T. A. 2005, A&A, 444, L9 MNRAS, 367, 433 Nesvadba, N. P. H., Lehnert, M. D., Eisenhauer, F., et al. 2006, ApJ, 650, Hernquist, L. 1990, ApJ, 356, 359 693 Hitomi Collaboration, Aharonian, F., Akamatsu, H., et al. 2016, Nature, Nulsen, P. E. J. 1986, MNRAS, 221, 377 535, 117 Nulsen, P. E. J., Powell, S. L., & Vikhlinin, A. 2010, ApJ, 722, 55 Hlavacek-Larrondo, J., Fabian, A. C., Edge, A. C., et al. 2013, MNRAS, Panagoulia, E. K., Fabian, A. C., & Sanders, J. S. 2013, MNRAS, 433, 3290 431, 1638 Peres, C. B., Fabian, A. C., Edge, A. C., et al. 1998, MNRAS, 298, 416 Hlavacek-Larrondo, J., McDonald, M., Benson, B. A., et al. 2015, ApJ, Pizzolato, F., & Soker, N. 2005, ApJ, 632, 821 805, 35 Rafferty, D. A., McNamara, B. R., Nulsen, P. E. J., & Wise, M. W. 2006, Iono, D., Wilson, C. D., Takakuwa, S., et al. 2007, ApJ, 659, 283 ApJ, 652, 216 Johnstone, R. M., Fabian, A. C., & Nulsen, P. E. J. 1987, MNRAS, 224, 75 Ruel, J., Bazin, G., Bayliss, M., et al. 2014, ApJ, 792, 45 Kirkpatrick, C. C., McNamara, B. R., Rafferty, D. A., et al. 2009, ApJ, 697, Rupke, D. S. N., & Veilleux, S. 2011, ApJL, 729, L27 867 Russell, H. R., McNamara, B. R., Edge, A. C., et al. 2013, MNRAS, 432, Li, Y., & Bryan, G. L. 2014, ApJ, 789, 153 530 Lim, J., Ao, Y., & Dinh-V-Trung. 2008, ApJ, 672, 252 —. 2014, ApJ, 784, 78 Ma, C.-J., McNamara, B. R., & Nulsen, P. E. J. 2013, ApJ, 763, 63 Russell, H. R., McNamara, B. R., Fabian, A. C., et al. 2016, MNRAS, 458, Magorrian, J., Tremaine, S., Richstone, D., et al. 1998, AJ, 115, 2285 3134 Markwardt, C. B. 2009, in Astronomical Society of the Pacific Conference Salomé, P., & Combes, F. 2003, A&A, 412, 657 Series, Vol. 411, Astronomical Data Analysis Software and Systems Salomé, P., Combes, F., Revaz, Y., et al. 2011, A&A, 531, A85 XVIII, ed. D. A. Bohlender, D. Durand, & P. Dowler, 251 Salomé, P., Revaz, Y., Combes, F., et al. 2008, A&A, 483, 793 McDonald, M., Benson, B., Veilleux, S., Bautz, M. W., & Reichardt, C. L. Salomé, P., Combes, F., Edge, A. C., et al. 2006, A&A, 454, 437 2013a, ApJL, 765, L37 Silk, J., & Rees, M. J. 1998, A&A, 331, L1 McDonald, M., Bayliss, M., Benson, B. A., et al. 2012, Nature, 488, 349 Simionescu, A., Werner, N., Finoguenov, A., Böhringer, H., & Brüggen, McDonald, M., Benson, B. A., Vikhlinin, A., et al. 2013b, ApJ, 774, 23 M. 2008, A&A, 482, 97 McDonald, M., Swinbank, M., Edge, A. C., et al. 2014, ApJ, 784, 18 Solomon, P. M., Rivolo, A. R., Barrett, J., & Yahil, A. 1987, ApJ, 319, 730 McDonald, M., McNamara, B. R., van Weeren, R. J., et al. 2015, ApJ, 811, Solomon, P. M., & Vanden Bout, P. A. 2005, ARA&A, 43, 677 111 Tozzi, P., Gastaldello, F., Molendi, S., et al. 2015, A&A, 580, A6 McMullin, J. P., Waters, B., Schiebel, D., Young, W., & Golap, K. 2007, in Tremblay, G. R., Oonk, J. B. R., Combes, F., et al. 2016, Nature, 534, 218 Astronomical Society of the Pacific Conference Series, Vol. 376, Vantyghem, A. N., McNamara, B. R., Russell, H. R., et al. 2016, ArXiv Astronomical Data Analysis Software and Systems XVI, ed. R. A. Shaw, e-prints, arXiv:1610.00716 F. Hill, & D. J. Bell, 127 Voigt, L. M., & Fabian, A. C. 2004, MNRAS, 347, 1130 McNamara, B. R., & Nulsen, P. E. J. 2007, ARA&A, 45, 117 Voit, G. M., Meece, G., Li, Y., et al. 2016, ArXiv e-prints, —. 2012, New Journal of Physics, 14, 055023 arXiv:1607.02212 McNamara, B. R., Russell, H. R., Nulsen, P. E. J., et al. 2016, ApJ, 830, 79 Williamson, R., Benson, B. A., High, F. W., et al. 2011, ApJ, 738, 139 McNamara, B. R., Wise, M., Nulsen, P. E. J., et al. 2000, ApJL, 534, L135 Young, L. M., Bureau, M., Davis, T. A., et al. 2011, MNRAS, 414, 940 McNamara, B. R., Russell, H. R., Nulsen, P. E. J., et al. 2014, ApJ, 785, 44